Flexan/Blake-XTM-Arc-3B-V1
Full fine-tune vs microsoft/phi-2. Fine-tune: 95% of the weights changed, across 32 of 32 layers.
Tensors changed: Embeddings 0/1, Norms 0/66, MLP 64/128, Attention 128/256, lm_head 1/2
2 derivatives checked
- mradermacher/Blake-XTM-Arc-3B-V1-i1-GGUF: Quantization
- mradermacher/Blake-XTM-Arc-3B-V1-GGUF: Same weights, converted